An agent-based approach for dependable planning of production sequences in automated production systems

Abstract

One approach to achieve flexibility and dependability for the control of automated production systems (aPS) is agent-oriented software engineering (AOSE). In this paper, the modular decoupling of the supervisory control, most significantly the planning of production sequences and transfer routes, from the distributed real-time control of the plant resources is demonstrated by the use of agents. The resulting product management agent (PMA) represents the technical process of the manufactured product and conducts a discrete reasoning algorithm to derive appropriate production plans by the use of graph-search and also by interaction with the underlying resource agents (RA). It is shown, that for a given production system, dependable solutions are automatically generated in regard to a given product recipe. Further it is deduced, that the solutions are calculated and evaluated by the PMA within a deterministic time duration. This is argued on the fact, that the computation complexity does not exceed polynomial time and is mostly predetermined by the design parameters of the plant. Thus, it gives a reasonable approach for the use in a real-time environment. Additionally, through separation of supervisory and field control, a modular software engineering is achieved, offering the advantage that the PMA and the resource agents can be reused, by solely adapting the knowledge bases and without the need for modifying the planning algorithms after a reconfiguration of the aPS.

Cyber-physical production systems architecture based on multi-agent’s design pattern—comparison of selected approaches mapping four agent patterns

“…Design patterns based on MAS for manufacturing would enable rapid application in industry [17]. Besides, the agentbased architectural solutions usually possess the characteristic of "plug and produce" use and are applicable in many domains after simple parameter adjustment [19]. In addition, they are suitable for different control layers regarding the automation pyramid that will allow the creation of versatile approaches regarding CPPS [9].…”

“…The approach of Rehberger [19] is designed for achieving both flexibility and availability during run-time (for coping with unknown product recipes and breakdowns of sub Table 2 Description of characteristics for CPPS [8,12,20,21] Feature Description Flexibility/changeability [12,21] (FC) It is often the grade to which a product or system can be used with effectiveness, efficiency, freedom from risk, and satisfaction in contexts beyond those initially specified in the requirements Reliability [21] (RL) A set of attributes that bear on the capability of software to maintain its level of performance under stated conditions for a stated period of time (four attributes: maturity, fault tolerance, recoverability, reliability compliance)…”

“…The first architecture of Wannagat [27] concentrates on the field level control and presents a MAS architecture for hard real-time and dependability, applied to PLC controllers (the most popular in industrial environments). Based on that, many other authors continue to build their MAS architectures on it (Folmer [48], Schütz [28], Rehberger [19], and Ulewicz [7]). Second author Fischer [32] also applied MAS for hard real-time including control level to put adaptability, flexibility, and dependability attributes into MFS.…”

“…A Diagnosis Module detects failures within the sub-agent's status, which identifies the existing situation based on the sensor data (signals measurements or other sub-agents' messages from Agent Interaction). Table 6 shows the design pattern for the RA in production plants [19,28].…”

“…It contains the sub-agent called AMS, the DF, and the MTS, all of them from FIPA standard. These are used in the same manner as presented by Wannagat et al [19,28] in Section 4.1. Fischer's pattern focuses on the MFS, but both patterns (Fischer-Wannagat) are used for the real-time intelligent conveyors' rerouting.…”